Abstract: Social AI refers to conversational artificial intelligence systems designed to meet users' social needs. Currently, it is permeating users’ daily lives with a subtle yet rapid momentum. While enhancing people's quality of life, it also brings about potential risks such as inducing users to disclose private data, blurring privacy boundaries, misusing user data, and manipulating user decision-making. Consequently, ethical governance regarding responsibility for privacy protection in Social AI has become urgent. However, given characteristics such as multi-agents and non-intentionality inherent in Social AI, significant challenges arise in the allocation of moral responsibility. The concept of distributed responsibility for non-intentional agents posits that all participants are responsible entities, thereby addressing the issue of missing responsible subjects. Starting from a backward-propagating distributed responsibility allocation approach, this framework clarifies the retrospective responsibilities of each stakeholder. Furthermore, sharing overall responsibility through a distributed model helps incentivize enterprises and designers to enhance their prospective responsibilities, thereby preventing harm. Thus, applying responsibility ethics governance based on distributed responsibility offers a viable solution for addressing privacy protection issues in Social AI.